Melon’s Puzzle Zoo has ended

Around the time of this posting, LMI is no longer accepting score submissions for the contest. Below is a link to the puzzle booklet, the same as the one on LMI without the password. There is also a PDF that has all of the puzzles on the test together with their solutions (don’t open if you don’t want to see them yet!), as well as some brief comments from myself about the puzzles. Mostly I pointed out the tricky innovative logic the puzzle had, if any — and for the adults there almost always was some.

Puzzle PDF
Solution PDF

Some thoughts on the test from my angle, coming from shortly after it is over and before seeing other comments: All in all, a lot like my first WPC performance. Passable and with its high points, but my inexperience clearly showed through in other areas.

Probably very few will disagree when I say I think it was too hard, both for casual solvers and for those gunning for top spots. This is pretty clear from the fact that the top scorer (congrats Psyho) finished 16 out of 20 puzzles, whereas most LMI tests are intended be finished by top contestants at least a few minutes early. Mine too, key word being intended in this case. I do realize that such an unexpected level of difficulty can really throw people off, especially for people used to finishing or at least getting close, so I apologize for this. It was not my goal to have no one come even close to a perfect score.

Suffice it to say this was mostly my fault. There were two rounds of testsolving, but in both cases the drafts sent out were way over the top, and I was probably a bit too stubborn about making the babies easy enough or eliminating some of the more ridiculous steps in the adults when revising. The final version is one in which motris, one of the three testers, would have probably finished with a little time to spare — it’s hard to say exactly because you can’t get genuine times from tweaked puzzles if you’ve done the original. This apparently was not an optimal standard to match up with, perhaps because motris is on fire these days and he’s also more familiar with my kinds of puzzles than a typical world-class competitor.

Also, I think the testsolver reports were the first time in these 1+ years that I ever heard how long someone else took to do a puzzle of mine, since previously I had only heard qualitative descriptions. So seeing their times was a slap in the face that I failed to completely react to.

Yet another thing that I probably did not give due consideration was the types themselves. There were no classic types on the test, and many of them were far from ordinary. The preview series probably alleviated that at least a bit, but of course not everyone would have had time to work through it all.

Still, it was experience for me. I’m not sure I would get the challenge level of a second competition perfect, but I think I would do a lot better. We’ll see, if it happens.

On subjects not related to difficulty, one issue was some of the answer extraction. Castle Wall and International Borders didn’t have the best mechanisms, although a good portion of mistakes made were almost certainly caused by messing up the puzzle somewhere as opposed to goofing the extraction. I’m not sure what Castle Wall’s mechanism should have been, but International Borders maybe should have asked for the number of black squares in each row instead, which is much more visually obvious.

Anyways, as I’ve said in several places, the puzzles on the test are my very best work, even if I did overdo the level of difficulty a bit, so I encourage you to work out any you didn’t get time to do this weekend. The highlights in my opinion:

  • Alien 1 (Castle Wall) Adult
  • Mutant 2 (Line Nurikabe) Adult
  • Mutant 4 (Nonconsecutive Fillomino) Adult – my favorite by a long shot
  • Hybrid 2 (Yajilin / Akari) Adult

(No babies in the list, I know. They’re not bad puzzles, but I don’t think any were standouts.)


12 Responses to “Melon’s Puzzle Zoo has ended”

  1. Scott Handelman Says:

    I wasn’t able to get a 2 hour window open to participate this weekend, but I did manage to print out the puzzles and try them. I have finished 15 out of 20 thus far. The Double Back/Country Road printed weirdly for me, with part of the grid missing…I tried to fill in what I assumed the grid to look like and correctly solved it, but I apparently guessed wrong, so I’ll try to solve the actual puzzle tonight.

    I’ll comment more once I’ve finished all 20. I know I wouldn’t have finished many in the 2 hours given.

  2. mathgrant Says:

    Whoa, I alpha-tested this thing, and I didn’t see the word LIAR. Nice. 🙂

  3. rob Says:

    Thanks a lot for making this test. I really enjoyed those puzzles I worked on, even with the time pressure, and I’m looking forward to those I skipped or messed up — the nonconsecutive fillominos in particular from your comment above.

    A few comments on the difficulty. Note that I’m not an experienced competitive puzzler.

    First off, I think it shouldn’t be a target to have the top puzzlers finish a competition. It’s much better if the top places are decided by who was able to finish all or “that really hard akari ex” rather than who was five minutes faster than the next.

    Secondly, I think experience with your puzzles helps a lot. I did a lot better on this competition than on the few others I’ve tried, placing 11th. I solved more than half of the puzzles, which seems to be roughly as good or better than my usual performance. I expect that a top puzzler with a similar level of familiarity with your puzzles would have a good chance of getting close to completing the test. Since it’s possible (or might be expected) that some of the best competitors know your puzzles or prepare for such a test, the difficulty level seems not so far off to me.

    Regarding individual puzzles, the Baby Liar Slitherlink threw me off, like someone posted on the lmi forum. I didn’t find a proper break-in before giving up to try something else. I usually love your Liar Slitherlinks, so would have thought those would come to me more easily. At any rate, I look forward to doing them in the future.

    I thought the International Borders puzzles were quite easy. In particular, the Adult one didn’t seem to have any intricate logic. The puzzle type does feel a little more exotic than others, though, and probably benefits a lot from practice with it.

    I got through the Hybrid puzzle series more smoothly than expected. In particular, they seemed quite a bit easier on average than the other adults, which does appear to be slightly reflected in the scoring.

    Regarding the puzzles I didn’t do: I skipped the Out of Sight and Akari Ex since I’m not that comfortable with the types and I don’t usually enjoy them quite as much. I did try the Baby Akari Ex in the last few minutes but failed to break in. As remarked above I failed to break in to the Liar Slitherlink. For the Nonconsecutive Fillomino, I ran into a contradiction on the Baby about two thirds of the way through, then skipped the Adult. I should probably have tried doing more of the babies from a competitive perspective, since those give more points per time (if you don’t get stuck).

    Anyway, thanks a lot for the test. I personally didn’t see anything to criticize, but I still won’t be surprised if you do even better in the future. And feel free to take a break for a bit!


    • rob Says:

      Some updates on the other puzzles.

      The Adult Nonconsecutive Fillomino is truly great. It took me quite a while to solve: quite a while of not getting in at all, then trying an Arukone-style approach, finally doing the addition, an “intuitive” approach leading to a contradiction, and finally redoing it correctly from the start. Beautiful puzzle.

      I forgot to say before that I really liked the Adult Line Nurikabe. The way the those five-lines force an island to connect to the border is something traditional Nurikabe can’t do.

      I did the Baby Liar Slitherlink after reading your comment on the LMI forum — that is a break-in I should really have noticed. I don’t think I saw that technique in your puzzles before, so maybe that’s what threw me off. In the test, I was closer to breaking in to the Adult one. Just now, both took me roughly the same time to solve, just over 10 minutes.

      Nothing much to say about the Out of Sight puzzles. The Baby one really was easy (< 3 min), while the Adult one took me a while (10 min), but rather because the puzzle type is a bit uncomfortable for me.

      Finally, the Akari Ex. I had real trouble with the Baby one, getting stuck a couple of times (7 min). In the end, I think it's quite an elegant puzzle, with nice symmetry and surprisingly few clues. Then the scary Adult Akari Ex fell really easily: It took me a bit longer than the Baby (9 min), but that's just because of the size. It certainly helped that my Akari Ex experience was applicable here. In particular, having seen the embedded graph thing before.

      I'm sorry for making you read so much…


      • MellowMelon Says:

        No need to apologize; I’m happy to read all such feedback on these puzzles. It’s good to hear that the NC Fillomino adult and the Line Nurikabe adult were as well-received as I hoped they would be. Two cases where I was proud of the idea before having even implemented it – although for the Fillomino it was probably 20+ cumulative hours of time trying to make it more than just an idea, and I did need to cheat in the end.

  4. stigant Says:

    Looking forward to completing these at a leisurely pace. Don’t let the time thing bother you too much. There’s no real prize riding on the outcome of the contest.

    >>First off, I think it shouldn’t be a target to have the top puzzlers finish a competition. It’s much better if the top places are decided by who was able to finish all or “that really hard akari ex” rather than who was five minutes faster than the next.

    That sort of test means that the result may hinge on which competitor accidentally chose the correct combination of puzzles to tackle. It’s really not possible to accurately gauge how difficult a puzzle is by just looking at it. The preference among top-level solvers (I believe, since I’m not one of them) is that they’d rather have total time (which still measures relative skill) be the tie-breaker than have the result hinge on something completely random like the order in which you happened to tackle the problems.

  5. Thomas Snyder Says:

    Thanks to MellowMelon for putting together this test. I wanted to address a few comments from the post-test discussion, to highlight a few issues.

    Speaking as one of those top solvers, I would prefer to not have a test/round of a competition end with 50-60% completion so that getting fortunate with puzzle selection trumps solving skill when scoring is not highly accurate. I’d rather the winner of a USPC/WPC finishes 90%+ of the puzzles and has to solve fill-in puzzles and loop puzzles and math puzzles and observational puzzles rather than just choose a smaller number of sub-disciplines and get lucky that there were over-valued puzzles for them in that subset.

    But I also especially dislike rounds that eventually hinge on just one puzzle, which is a huge outlier from the rest, as happened during a team round at “WSC” 4. There, most teams had 9 of the 10 puzzles solved readily within the time (we had an hour left at that point) and then the 10th, a rather broken design with no logical path to a solution, took everyone the rest of the time with few finding a solution (we still had 9 of 10 solved after that last hour). Saying “Who solved that really hard akari ex” should be the winner is a similar problem to having too few puzzles determine a winner. Ideally a test should be finished so that everyone has passed the same set of obstacles and total time is indicative of the best performer, but if not then a ~90% completion rate is very acceptable.

    Fortunately, I think this test worked alright, if still a bit too hard, so that the 80% result by psyho, while a little below the goal 90+% I have in mind, seems fair. My testing times would disagree with the comments that the Akari EX was an impossibly hard puzzle or the Baby Liar Slitherlink was also an outlier puzzle when the difficulty when this whole test had “outlier” puzzles in the sense that the logical steps were new or framed very differently than past experience.

    If anything, I would surmise the Akari EX was a less friendly/familiar type so solvers didn’t spend time on it, or if they did they were less willing to experiment rather than deduce to catch onto the solution path. I worked out the middle of the Akari EX very much be seeing the results of some guesses, which I am willing to do on such an Akari in competition. Given how different solvers approach different puzzles, the times will vary a lot on a test like this when so many puzzles require original deduction ideas of such varied flavors that some solvers will get some and others will see others but its very difficult to predict what is a “fair” score for any individual example.

    As I said, I test-solved the test. I actually saw a super-sized version with 30 puzzles (5 groups in each type, Baby and Adult of each of those) and after finishing it in 3 hours my comment was the test “destroyed me”. Certainly, expecting a roughly 2 hour experience, the extra hour of content was too much even for me to get through. I think Melon took the right steps to end up with a good test, but when the last legit test-solve is on something that is fairly different than the actual test presentation, making the right level of changes will be tough.

    I really liked the whole set of puzzles Melon put together here. The NC Fillomino was tremendous, even if it took the longest of all the puzzles for me, and even the 10 puzzles that were cut from the version I tested were very good. My conclusion is that it’s a tremendous group of challenges without a clock, but perhaps in the final analysis a bit too hard within a fixed time limit for people to have the best experience with them. I have my own opinions on what are good competition puzzles and what are good sit-down-and-think puzzles, and I don’t know that the two overlapped, but I was quite intrigued by everything MellowMelon put in this test.

  6. Andy Says:

    The dread I felt earlier turned to excitement as I started the test – those 2 hours went quickly! I had a blast, it was a very enjoyable test, both during the competition, and afterwards as I went back to the ones I didn’t have time to complete. I finished well behind the superhumans, of course, but was pretty happy overall with my results – all the time I’ve invested in enjoying this site helped a huge amount.

    Like everyone I’m sure, I’ve got a few regrets – afterwards I made it through the Nurikabe/Fillomino & Double Back/Country Road hybrids fairly quickly, I wish I’d attempted those (and the baby IB I didn’t even look at for some reason) instead of investing too much unsuccessful time on the baby NC Fillomino (got stuck only to afterwards find an easy clue I’d completely overlooked), adult Line Nurikabe, and Yajilin/Akari hybrid (stuck at 3/4 finished).

    From the viewpoint of a more average, non-contending solver, the overall difficulty didn’t bother me at all. I didn’t expect to come anywhere near finishing everything, completed most every puzzle I tried, and was left with a definite sense of accomplishment even though I wish I’d chosen puzzles differently. A few notes:

    I was surprised to see some complaints in the discussion thread regarding the baby Liar Slitherlink, I guess I was lucky to see the break-in right away – it seemed to be highlighted, put in the center rows, to force valid clues in the center columns, and everything fell from there. I didn’t try the adult until after, but I liked that one as well, saw the break-in corners quickly and had a much slower, but nice solve. IMO both are fair puzzles for the competition.

    I got most of the Yajilin/Akari during the test, except for the lower left quadrant.. Afterwards I was able to painstakingly exempt light possibilities square-by-square to find the eventual solution, but based on the high number of successful completions, I must have missed some logical insight there.

    It was interesting to see your note about the explicitly marked 2 in the Double Back/Country Road, that it was meant to highlight the break-in. This actually caused some problems for me, since I was really surprised to see a spurious clue (which wasn’t inserted to fulfill some symmetry/aesthetic aspect) – it made me stop for a while to wonder if I had the rules wrong, it seemed to imply that it was possible for the path to *not* go through a 2-square region. But again, I really love both of these types, and their hybridization.

    I was glad to see your note about the adult NC Fillomino, it never would have occurred to me to count the total region size, I look forward to completing the puzzle using that realization. My only other remaining unsolved puzzles are the adult Line Nurikabe and particularly both Akari EX – I must have a real problem with those, I’ve stared at the baby for probably over 20 minutes without seeing the break-in..

    Thanks again! I really had a great time with both the intense time-pressured test duration, and the relaxing post-test cleanup..

    • MellowMelon Says:

      Thank you for the comments on the puzzles.

      Glad to see the Baby Liar Slitherlink break-in wasn’t completely missed, but given the solving statistics it was apparently not an easy find regardless.

      For the Yajilin/Akari, I will confess finishing the lower left logically is possible but quite involved. This was a case where it’s something I ordinarily would have tweaked to be better, but that type is just… it was hard enough getting a unique solution at all. Truthfully I was kind of surprised by the solving statistics too. I wonder if there’s a good method to finish it that I missed, or if it was just that easy to bifurcate once you were that far.

      Adult Line Nurikabe and Adult Akari EX have nontrivial insights behind them that you might enjoy cracking on your own. But if you do get stuck the solutions PDF will spoil both. As for the baby Akari EX, that one is a short sequence of tricky logic, and it was slotted as a baby (perhaps inappropriately) because 10 by 10 Akari are insanely easy to brute force.

      • Jack Says:

        For the lower-left of the Yajilin/Akari, my solving path went something like this (trickier than the rest, but not that involved):

        There must be a lightbulb somewhere in c1, and with a little work you can narrow it down to either r7 or r9. But similarly, c5 also has a bulb in either r7 or r9, since the path segment sticking in from the right can’t hit both of those corners (well, they’re corners by the time you get to that part of the puzzle). That means there are no bulbs in between in either r7 or r9. The lack of lightbulb in r7c3 means the only way to light up r6c3 is from r6c4. The bulb there forces a bulb in r10c2. The path coming up out of the bottom row can’t go through the corner in r9c5, since it would strand r8c4 (already lit up, no way to put the path through it). So r9c5 is a bulb and r7c1 fills in to match. Then you just draw the path.

        My favorite of the Baby puzzles was probably the Akari EX — very clever break-in, with just enough information from the center to break the symmetry and force the solution. I caught the nice break-in for the Baby Liar Slitherlink, but managed to break the puzzle somewhere else during the contest time. I didn’t think it was a particularly “mean” puzzle though. Still finding the time to do some of the adult puzzles…

      • MellowMelon Says:

        Actually, that was the best logical method I ever came up with for that last corner too. Maybe it was easier to find than I thought, although I’m willing to bet several people just saw their way to the end.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: